Yao Meta Skill

SkillAI & models

Lets your agent create, improve, and package reusable skills from prompts, docs, and notes.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Yao Meta Skill skill

About this capability

Create, improve, or evaluate an existing skill from workflows, prompts, SOPs, scripts. Use for migration/release/package, routing, evals/tests, install/trust checks, 优化已有 skill, 补 trigger 评测. Exclude summary/translation/docs, brainstorming, one-off prompts, copy-only edits, outline-only requests.

What this skill tells your AI

The instructions your AI receives, as published by yaojingang/yao-meta-skill in SKILL.md and read by ahel’s review.

Router Rules

  • Route by frontmatter description using the lightest mode.
  • Keep SKILL.md lean: guidance in references/, logic in scripts/, evidence in reports/.

Modes

  • Scaffold: exploratory; Production: team; Library: shared; Governed: release-critical.
  • Rules: Method, Modes, Boundaries.

Compact Workflow

  1. Require repeated use + reusable output contract; skip one-offs.
  2. Lock target path + identity before writes; self-edits require --self.
  3. Capture job, output, exclusions, constraints, standards, mode.
  4. Scan 3-5 useful external/user/local references.
  5. Write description early; route edits need trigger_eval.py; release gates follow risk.
  6. Add deeper profiles only when earned.

Playbooks: Target Safety, Intent, Skill IR, Output Eval, Review Studio.

Updates

Activation: scripts/yao.py check-update --notice --self; show notice_text if requested; continue. “更新”: scripts/yao.py self-update --self --yes.

Skill OS 2.0 Gates

For production/library/governed releases, run Skill IR, compiler, trigger/output eval, Atlas, conformance, trust, package/install, upgrade, drift, waiver, Review Studio.

Governed Package Boundary

Governed/file-backed packages: label input_files a file-backed fixture; include owner, review cadence, output contract, rollback boundary; require trust report and reports/output_quality_scorecard.md; mark unavailable telemetry/approvals/metrics/benchmarks as missing evidence; never fabricate evidence.

Preserve labels literally when they apply: file-backed fixture, input_files, output contract, rollback boundary, trust report, reports/output_quality_scorecard.md, missing evidence.

First-Turn Style

  • Infer non-core gaps; ask one core-fork question per round, max two, then use recorded preferred inference.
  • In Chinese, sound soft and companion-like; use Intent Dialogue.

Output Contract

Builds: SKILL.md, aligned interface, justified assets, boundary/gate summary. Audits: findings + fixes; edit only when asked. No-skill: no files.

Reference Map

Primary: Method, Artifact Design, Systems, Governance, SkillOps.

Signals

GitHub stars
3k
Forks
244
Last commit
Aug 2026
Advanced
Catalog kind
skill
Gateway key
yao-meta-skill
Source
github.com/yaojingang/yao-meta-skill